Pre-Main Sequence Accretion Rates at Low Metallicity 3

Draft version December 5, 2014 Preprint typeset using LATEX style emulateapj v. 05/12/14

PRE-MAIN SEQUENCE ACCRETION IN THE LOW METALLICITY GALACTIC STAR-FORMING REGION Sh 2-284 V. M. Kalari & J. S. Vink Armagh Observatory, College Hill, Armagh, BT61 9DG, UK; [email protected] Draft version December 5, 2014

ABSTRACT We present optical spectra of pre-main sequence (PMS) candidates around the Hα region taken with the Southern African Large Telescope, SALT, in the low metallicity (Z) Galactic region Sh 2-284, which includes the open cluster Dolidze 25 with an atypical low metallicity of Z 1/5 Z . It has been suggested on the basis of both theory and observations that PMS mass-accretion∼ rates, M˙ acc, are a function of Z. We present the first sample of spectroscopic estimates of mass-accretion rates for PMS stars in any low-Z star-forming region. Our data-set was enlarged with literature data of Hα emission in intermediate-resolution R-band spectroscopy. Our total sample includes 24 objects spanning a mass range between 1 - 2 M and with a median age of approximately 3.5 Myr. The vast majority (21 out of 24) show evidence for a circumstellar disk on the basis of 2MASS and Spitzer infrared photometry. We find M˙ acc in the 1 - 2 M interval to depend quasi-quadratically on 2.4 0.35 0.7 0.4 stellar mass, with M˙ acc M ± , and inversely with stellar age M˙ acc t− ± . Furthermore, ∝ ∗ ∝ ∗ we compare our spectroscopic M˙ acc measurements with solar Z Galactic PMS stars in the same mass range, but, surprisingly find no evidence for a systematic change in M˙ acc with Z. We show that literature accretion-rate studies are influenced by detection limits, and we suggest that M˙ acc may be controlled by factors other than Z , M , and age. ∗ ∗ Subject headings: stars: pre main-sequence, stars:variables: T Tauri, Herbig Ae/Be, accretion disks, stars: formation, stars: fundamental parameters

1. INTRODUCTION (2005) using similar techniques, where the authors noted Present understanding suggests that most stars accrete that these stars also exhibited Hα emission and were lo- mass from a circumstellar disk during their pre-main se- cated in H II regions. These studies showed that the lu- quence (PMS) phase (see Hartmann 2008). The rate minosities of these objects were significantly higher than ˙ those of Galactic PMS stars at solar Z , and with similar of mass accretion (Macc) from the disk to the star is spectral types. de Wit et al. (2003) showed that HAeBe vital to describe the system’s evolution, including the candidates in the Small Magellanic cloud (SMC) at lower potential growth of planets in the disk whilst the star metallicities (Z 0.002 - 0.004) were even brighter than reaches its main sequence configuration. Viscous disk their LMC counterparts.∼ Based on the inferred lumi- ˙ evolution models predict that Macc drops with time as nosities of the Magellanic clouds (MCs) PMS stars, de ˙ 1.5 Macc t − , a prediction which has been substantiated Wit et al. (2005) suggested that the protostellar mass- by empirical∝ ∗ studies (Hartmann et al. 1998). Observa- accretion rates are inversely proportional to metallicity, tions have also found that the mass-accretion rate scales leading to more luminous PMS stars that are visible at ˙ α with stellar mass as Macc M over a mass range from younger ages at lower metallicities. Whether the PMS 0.3 M -3 M , with index α∝ 2∗ (Calvet et al. 2004; Natta M˙ acc are analogously related to metallicity is not clear. et al. 2006; Muzerolle et al.∼ 2005; Sicilia-Aguilar et al. The M˙ acc estimated by de Wit et al. (2005) for their 2006; Manara et al. 2012), though it has been found to be best LMC PMS candidate (ELHC 7) indicated no signif- arXiv:1412.2014v1 [astro-ph.SR] 5 Dec 2014 discrepant at M < 0.8 M (Fang et al. 2009; Barentsen icant difference in comparison to the canonical Galactic ∗ ˙ et al. 2011). Note that the Macc vs. Mass relation is not average. reproduced by the current paradigm of disk evolution, On the contrary, studies based on the Balmer contin- and it is subject to ongoing research (e.g. Hartmann et uum excess (Romaniello et al. 2004) and Hα emission al. 2006). (Panagia et al. 2000) inferred higher M˙ compared to ˙ acc It has also been suggested that Macc is a function of the Galactic average in a LMC region close to SN1987A. metallicity, Z. Seven PMS star candidates discovered Follow-up Hα photometric studies by the same group in in the Large Magellanic Cloud (LMC) at Z 0.006 by ≈ the LMC regions around SN1987A (De Marchi et al. Beaulieu et al. (1996) were the first PMS candidates 2010) and 30 Doradus (De Marchi et al. 2011a) which identified in a low-Z region. Beaulieu et al. (1996) se- were summarized in Spezzi et al. (2012); and in the lected these objects because they exhibited peculiar pho- SMC open clusters NGC 346 (Z 0.002; De Marchi et tometric variability similar to Galactic Herbig HAe/Be al. 2011b) and NGC 602 (Z 0.004;∼ De Marchi et al. stars. The number of low-Z PMS candidates was in- ∼ 2013) reported a metallicity dependence on M˙ acc (see creased in the follow-up works of Lamers et al. (1999); de ˙ 1 ˙ 0.36 Wit et al. (2002); de Wit et al. (2003) and de Wit et al. below) whilst also finding Macc M , and Macc t in the 1-2 M range for the LMC∝ PMS∗ stars. They∝ ∗ em-

2 V. M. Kalari & J. S. Vink ployed a novel method using VI Hα photometry, rather suggesting that it is challenging for accretion to proceed than conventional spectroscopy to derive M˙ acc. We have at significant rates at older ages in low Z environments. applied a similar method in the Galactic regions IC 1396 It also suggests that M˙ acc may be higher at lower Z, but (Barentsen et al. 2011), NGC 2264 (Barentsen et al. only at young ages. 2013) and NGC 6530 (Kalari et al. 2014, in prep.), and It is thus essential to further scrutinize the differences found that the results from the photometric method do in M˙ acc of metal-poor PMS stars in comparison to solar not differ significantly from spectroscopic Hα or U-band Z PMS stars, as to understand star formation in different measurements, nor did we find a large number of inter- Z environments, as well as the processes involved in disk lopers. Furthermore, we discovered an accreting PMS evolution. To this purpose, we employ spectroscopic ob- candidate star in the LMC region 30 Doradus (Kalari servations of candidate PMS stars located in the Galactic et al. 2014), which is within 10 pc to one of the re- star-forming region Sh 2-284 (Sharpless 1959). Sh 2-284 gions examined in Spezzi et al.∼ (2012). These vari- is an H II region encompassing the central open cluster ous results suggest there may well be genuine highly- Dolidze 25 (also OCL-537). The B stars of Dolidze 25 accreting PMS stars present in the MCs. De Marchi et have been shown to be metal deficient with respect to al. (2013) suggest accretion rates are inversely propor- solar values, with all measured elemental abundances tional to metallicity based on studies of the two SMC ranging from 0.6 to 0.9 dex, leading to an approxi- open clusters. De Marchi et al. (2011b) reported a mate metal deficiency− of− 0.7 dex (Lennon et al. 1990; 8 1 − median M˙ acc of 4 10− M yr− in NGC 346 for a bi- Fitzsimmons et al. 1992). The association of Dolidze 25 modal age distribution× of 1 and 20 Myr; and De Marchi with the Sh 2-284 indicates that the Sh 2-284 region is 8 1 et al. (2013) found a median of 2.5 10− M yr− for also characterized by low metallicity (Puga et al. 2009; × 9 1 Delgado et al. 2010; Cusano et al. 2011). a 2 Myr population in NGC 602 and 4 10− M yr− for a 20 Myr population. De Marchi et× al. (2013) sum- Sh 2-284 is located at a distance of approximately 4 kpc marized their work in the MCs using an equation of the (Delgado et al. 2010; Cusano et al 2010) and provides a unique opportunity to observe PMS stars in the 1 - form log Macc = a log t + b log M + c, where a, b and c are constants, and× c is∗ inversely× proportional∗ to Z. Sim- 2 M range with medium-resolution spectroscopy, and we can compare our measured accretion rates with liter- ilarly, Spezzi et al. (2012) suggest that M˙ acc is inversely ˙ ature spectroscopic measurements of solar Z PMS stars, proportional to Z, and suggest that Macc in the LMC is as well as results claimed from Hα photometry in the higher than that of Galactic PMS stars having similar LMC by Spezzi et al. (2012) 1. masses, irrespective of age. Spezzi et al. (2012) found The paper is organized as follows. Section 2 details ˙ 8 1 a median Macc 5 10− M yr− of LMC PMS stars the data sample, and measured properties of the PMS having an average∼ age× of 10 Myr, when accretion in most stars. Section 3 describes our results and method to cal- low-mass (< 0.8 M ) Galactic PMS stars is thought to culate M˙ acc. A discussion of the M˙ acc distribution with have ceased (Fedele et al. 2010). Approximately, for 1 - respect to M , t and Z is presented in Section 4. The 2 M the PMS lifetime is between 30 - 5 Myr respectively conclusions of∗ our∗ work are summarized in Section 5. (Palla & Stahler 1999). This suggests that accretion is a dominant process for a large fraction of the PMS life- 2. DATA time in the LMC, contrary to current observations in For our program, we selected PMS candidates on the the Galaxy. Prolonged accretion at these rates suggests basis of infrared 2MASS and Spitzer photometry. Six accretion adds significant mass to the central star after Spitzer identified Class II sources Young Stellar Objects the ’birthline’ (Stahler 1983; Palla & Stahler 1990; Hart- (YSO) from Puga et al. (2009) with near-infrared JHKs mann et al. 1997; Tout et al. 1999) – contrary to current magnitudes from 2MASS (Two Micron all sky survey; disk or spherical accretion models. The authors suggest Cutri et al. 2003) or optical UBV magnitudes from Del- that the lower disk opacity, and viscosity due to lowering gado et al. (2010) were selected. This information is pre- Z leads to a longer viscous timescale, thereby allowing sented in Table 1. The location of the selected stars is accretion to take place at such older ages. overlaid on an Hα gray-scale in Fig. 1. Long-slit spectra However, observational studies of disk lifetime at lower of these objects were obtained using the Robert Stobie Z indicate that metal-poor stars in the extreme outer Spectrograph (RSS) on the 10 metre Southern African Galaxy (Z 0.002) lose their disks significantly more Large Telescope (SALT) located in Sutherland, South ≈ quickly than solar-Z PMS stars (Yasui et al. 2009). Ac- Africa (Burgh et al. 2003). A slit width of 1.2500 was used cretion at the high rates measured in the MCs by Spezzi to obtain good sky subtraction. A red grating covering et al.(2012) and De Marchi et al.(2010;2011a;2011b;2013) λλ 5700 - 7500 A˚ with resolution R 2000, and disper- means that the disks of these PMS stars have under- sion 1 A˚ per pixel was used. Observations∼ were obtained gone gravitational instability (Pringle 1981; Hartmann over 3 nights during December 2013. The seeing varied 2006; Cai et al. 2006), shortening the disk lifetime. Fur- between 1.1 and 1.500. thermore, lowering Z lowers the dust shielding efficiency Bias subtraction, flat-field correction, and cosmic ray against UV photoevaporation from OB stars leading to removal were performed using standard IRAF2 proce- possible disk erosion, an effect which has been demonstrated in the LMC (Spezzi et al. 2012). Internal pho- 1 Due to the lack of publicly available data of accretion properties toevaporation is also thought to reduce disk lifetimes at of PMS stars in the SMC, we restrict our comparison to the LMC PMS sample of Spezzi et al. (2012). low-Z, as the lower metal content increases X-ray opac- 2 IRAF is distributed by the National Optical Astronomy Ob- ity (Ercolano & Clarke 2010). In all these cases disk life- servatory, which is operated by the Association of Universities for times in low-Z environments are thought to be reduced, Research in Astronomy, Inc., under cooperative agreement with the National Science Foundation. Pre-main sequence accretion rates at low metallicity 3

TABLE 1 Literature data for observed stars

2MASS α (J2000) δ (J2000) VB VU BJJ HJ Ks D10a P09b − − − −

06443682+0016186 06 44 36.8 +00 16 18.70 ...... 14.58 0.74 1.54 ... 17 06445577+0013168 06 44 55.8 +00 13 17.30 15.23 0.32 0.65 13.89 0.43 0.99 307 37 06445837+0014151 06 44 58.4 +00 14 15.50 16.68 0.38 1.08 14.47 0.55 0.80 585 44 06450208+0019443 06 45 02.1 +00 19 44.00 ...... 14.62 0.49 0.78 ... 50 06450681+0013535 06 45 06.8 +00 13 54.00 16.59 0.61 1.13 13.95 0.77 1.57 1563 52 06451007+0014117 06 45 10.1 +00 14 12.10 15.71 0.58 0.91 13.37 0.00 1.61 1897 55

a Identiﬁcation no. in Delgado et al. (2010). b Identiﬁcation no. in Puga et al. (2009). 15

30'00.0" J06443682+0016186

24'00.0"

J06445837+0014151

5 18'00.0" Dec (J2000) Dol 25 J06451007+0014117 Normalized Relative ﬂux

12'00.0" 0

5800 6000 6200 6400 6600 6800 7000 7200 +0°06'00.0" Wavelength (A)˚

0.15 ◦ Fig. 2.— Normalized medium resolution longslit spectra of PMS 46m00.00s 40.00s 20.00s 45m00.00s 40.00s 6h44m20.00s RA (J2000) candidates in Sh 2-284 observed using the SALT telescope. From top to bottom we show the spectra of an A9, F0 spectral type PMS Fig. 1.— Sh 2-284 in Hα taken from Parker et al. (2005). North stars, and a F6 star with no Hα emission but infrared-excess. For is up and east is to the left. Stars observed with SALT are marked the complete results see Table 2. by circles, and stars with VIMOS spectra are marked by crosses. The central cluster Dolidze 25 is also marked. At d = 4000 pc, 0.15◦ corresponds to 10 pc. ∼ dures. Wavelength calibration was performed using Ne J06450681+0013535 J06445577+0013168 J06450208+0019443 arc spectra using the IRAF identify task. Spectra were 1.0 1.0 1.0 corrected for sensitivity using the Sutherland extinction 0.8 0.8 0.8 curve. The variable pupil design of the SALT telescope 0.6 0.6 0.6 makes absolute ﬂux calibration impossible. The ﬁnal ex- tracted 1-dimensional spectra are shown in Fig 2. The 0.4 0.4 0.4 average signal to noise is 20 - 40. 0.2 0.2 0.2

∼ Normalized Intensity 0.0 0.0 0.0 2.1. Basic analysis of the SALT data 0.2 J06445837+0014151 J06443682+0016186 J06451007+0014117 Equivalent widths (EW) of emission and absorption 1.0 1.0 0.0 0.8 0.8 0.2 lines were measured using the IRAF task splot. Spec- − tral types were derived from the EW of the (i) Mg II 0.6 0.6 0.4 − 5711 A(ii)˚ Mn II 6015 A(iii)˚ Ca I 5711 A˚ lines following 0.4 0.4 0.6 the method devised by Hernándezet al. (2004). Effec- − 0.2 0.2 0.8 tive temperatures (Teff ) were estimated using the spectral −

Normalized Intensity 0.0 0.0 1.0 type-Teff calibration of Kenyon & Hartmann (1995). Hα − 400 200 0 200 400 400 200 0 200 400 400 200 0 200 400 line profiles are displayed in Fig. 3. Five stars show Hα − − − − − − 1 in emission. Combined with the fact that these stars all Velocity (km s− ) show near to mid-infrared excess resembling PMS stars, Fig. 3.— Normalized Hα line profiles of the SALT spectra. The 1 we take these five stars to be bonafide accreting PMS velocities of all stars with Hα in emission is larger than 270 km s− , stars. indicating clearly accretion in the line profile. The measured Hα Model colors and bolometric corrections were calcu- EW are given in Table 2. lated using Castelli & Kurucz (2004) model spectra at 4 V. M. Kalari & J. S. Vink

3.0 the appropriate Teﬀ and Z. We assumed the logarithm of surface gravity log g = 4.0 1.0. Where available, BV ± photometry from Delgado et al. (2010) was used to 2.5 determine the reddening E(B V ) and the logarithm 0.3 Myr of luminosity (log L). We used− a reddening law with 1 Myr 2.0 RV = 3.1 to determine the absolute visual extinction AV . 3 Myr We adopted a distance d = 4000 400 pc derived by Cu- ZAMS ± 3 M sano et al. (2011) and Delgado et al. (2010) for cal- 1.5 culating the luminosity. Two stars (J06445578+001368 L/L 2 M

and J06450208+0019443) that do not have Delgado et al. log 1.5 M 1.0 (2010) optical photometry, archival B V colors (Monet − 0.8 M et al. 2003), and 2MASS J-band magnitudes were used to derive the extinction and luminosity respectively. We 0.5 assumed that the effect of any infrared excess in the J- band is comparatively small. Table 2 summarizes the 0.0 10 Myr fundamental parameters determined for these stars using SALT spectra. 4.2 4.1 4.0 3.9 3.8 3.7 3.6 3.5 log Teff 2.2. VLT/VIMOS data Fig. 4.— Hertzsprung-Russell diagram of Sh 2-284 PMS stars. Additional data of PMS stars in Sh 2-284 was provided The solid vertical lines from right to left are Z = 0.004 PMS in Cusano et al. (2011). These authors obtained mul- isochrones (Bressan et al. 2012) at 0.3, 1, 3 and 10 Myr, and the zero-age main sequence respectively. The dashed horizontal tislit spectroscopy of 900 objects covering a 30 30 lines from bottom up are the 0.8, 1.5, 2, and 3 M tracks. The 2 ∼ × arcmin area (see Fig. 1) using the Visual & Multi- average error bars in log Teff and log L are shown in the top-right Object Spectrograph (VIMOS) on the Very Large Tele- hand corner. The log Teff and log L of the majority of our PMS stars have been taken from Cusano et al. (2011). See Table 3 for scope (VLT). They identified 23 bonafide PMS stars from the interpolated ages, and masses. this sample based on Hα emission, spectral type, and near-infrared excess criteria. 17 show infrared-excess re- Russell (H-R) diagram Fig. 4. Overlaid are the non- sembling Class II sources, whilst the remaining objects accreting single star isochrones and tracks calculated by resemble Class III sources (i.e. weakly accreting PMS Bressan et al. (2012) at Z = 0.004. The masses and stars). The location of these stars is shown in Fig. 1. ages are measured by interpolating their positions with Three stars (J06443682+0016186, J06450208+0019443, respect to tracks and isochrones, respectively. The de- J06450681+0013535) also have SALT spectra. rived masses and ages are listed in Table 3. The differ- The spectral types and fundamental parameters for 22 ences that arise due to variations between different sets stars were determined by Cusano et al. (2011) using of tracks and isochrones are discussed in Appendix A. the method devised by Hernández et al. (2004). One The mass range (0.9 - 2.6 M ) places a majority of the star (J06452476+0013360) shows no absorption features stars in the intermediate mass T Tauri or Herbig Ae type which means that it did not allow for spectral classi- range. These stars are very similar to accreting Classi- fication. We adopt these results which are listed in Ta- cal T Tauri stars and their M˙ can be reliably mea- ble 2. We compared the spectral parameters for the three acc sured using their Hα emission line luminosity (Calvet et stars in common with our SALT observations, and we al. 2004). The median error on the masses is 0.1 M . found that spectral types agreed within 2 sub-classes, . The ages determined from isochrones span a range of 1 - which is the average error. The agreement is expected 15 Myr. The PMS stars in the central cluster Dolidze 25 as both spectral type determinations use similar line in- have ages around 2 - 3 Myr, including the newly identified dicators and distances adopted by the authors are the PMS stars using the SALT spectra (see Fig. 1). Those same as the values used in the analysis of the SALT located at the edges of the surrounding nebulous bubbles spectra. The logarithm of luminosity of one overlapping (see Puga et al. 2009 for a detailed description of the star (J06443682+0016186) measured using near-infrared nomenclature) have younger ages, whereas those at the J-band magnitude is higher by 0.25 dex to the value mea- rim have older ages. These results are in agreement with sured using Cusano et al. (2011) photometry. This is the near-mid infrared study of Puga et al. (2009) and attributed to the near-infrared excess. For this star, we spectroscopic-photometric work of Cusano et al. (2011). adopt the logarithm of luminosity determined by Cu- Puga et al. (2009) interpret this as evidence that star sano et al. For all other parameters determined using formation was triggered by a previous generation of ion- both SALT and VIMOS spectra, we adopt the values izing stars in these regions, which has lead to the spread estimated from the SALT spectra. of PMS ages inferred from the H-R diagram. Their ar- The overall sample consists of 24 bonafide PMS stars in gument was supported by a first-principle analysis of the Sh 2-284 with spectral types ranging from mid A- early gas dynamics within the region. G for which there are observed parameters, which are PMS membership of these stars is not suspect, as all necessary to infer accretion properties. objects display infrared excesses resembling PMS stars, 3. RESULTS and there is little evidence for mid-A to early-G main sequence stars having Hα EW < 5 A.˚ The older stars are 3.1. Masses and Ages located in the nebulous B2 and− B3 regions (see Fig. 1). Using the measured effective temperatures and lumi- Therefore, the possibility that they are background stars nosities, we place the PMS stars in an Hertzsprung- detected through nebulosity is small, as they would have Pre-main sequence accretion rates at low metallicity 5

TABLE 2 Physical parameters of PMS stars in Dolidze 25

a b ID (2MASS) Spectral EW (Hα) AV log Teﬀ log L type (A)˚ (mag) (K) (L )

SALT spectra J06443682+0016186 A9 8.98 0.4 1.62 0.25 3.87 1.10c J06445577+0013168 F2 −8.23 ± 0.3 1.93 ± 0.15 3.84 1.40 J06445837+0014151 F0 −26.58± 1.3 2.45 ± 0.15 3.85 1.38 J06450208+0019443 A6 −43.56 ± 2.1 2.23 ± 0.30 3.92 1.53c J06450681+0013535 F9 − 1.46 ±0.5 2.06 ± 0.15 3.78 1.26 − ± ± VIMOS spectrad J06443682+0016186e A8V 8.4 0.2 1.9 3.88 0.98 J06450208+0019443e A5V −8.05± 0.1 2.44 3.91 1.27 J06450681+0013535e G0V − 1.9 ±0.1 2.0 3.77 1.25 J06451701+0022077 A0V −10.3± 0.4 3.82 3.98 1.48 J06450971+0014121 A5V −14.0 ± 0.2 2.11 3.91 1.83 J06443291+0023546 F0V −23.8 ± 0.4 3.1 3.86 1.27 J06443788+0021509 F0V − 2.8 ±0.1 2.28 3.86 1.27 J06443541+0019093 F2V −4.2 ± 0.2 3.0 3.83 1.0 J06443827+0019229 F2V −11.7 ± 0.25 1.3 3.83 0.66 J06451727+0023454 F2V −16.58± 0.3 2.69 3.83 1.0 J06450274+0018077 F3V −33.7 ±0.25 3.14 3.84 0.74 J06451616+0022238 F3V − 5.3 ±0.25 3.01 3.84 0.53 J06443294+0010528 F4V −139.28± 1.23 3.64 3.81 1.11 J06444496+0019335 F8V − 8.8 ±0.15 2.52 3.78 0.88 J06444602+0019182f G0V −73.2± 0.9 5.63 3.77 0.7 J06451318+0018307 G0V −40.5 ± 0.5 2.64 3.77 0.29 J06451841+0022189 G4V − 5.2 ±0.3 1.05 3.75 0.19 J06444936+0020245 G5V −14.87± 0.3 1.7 3.75 0.42 J06452454+0022448 G8V −21.18 ± 0.2 2.33 3.73 0.75 J06450075+0013356 K0V −31.69 ± 0.5 1.89 3.69 0.28 J06444714+0013320f K2V − 20.5 ±0.5 2.15 3.68 0.36 J06452774+0015514 K4V −29.0 ± 1.3 2.77 3.64 0.06 − ±

a a Average error in log Teff is around . 0.05 dex due to uncertainty in spectral class. The spectral class classification is uncertain within 1.5 subclasses from SALT spectra, and 1 subclass from VIMOS spectra. b Average error in log L is 0.15 dex mainly due to uncertainty in d 500 pc. c Log L was calculated using∼ 2MASS J-band magnitude and archival∼ (Monet et al. 2003) B V colours. d Taken− from Table 4 in Cusano et al. (2010), where the quoted average errors in spectral type is one subclass, AV 0.1 mag. e Star also has SALT longslit∼ spectra. f Star not detected in 2MASS, coordinates from Cusano et al. (2010). been significantly more extincted than the other PMS (Tognelli et al. 2011) used for age determination by Cu- stars. Moreover, one early G star (J06451318+0018307), sano et al. (2011) in Appendix A. We find that both with an estimated age > 10 Myr, also displays Li in ab- isochrones gives similar absolute age values. Overall, we sorption in high signal-to-noise high resolution VIMOS take . 1 Myr as an average error for the interpolated age. spectra. Li λ 6708 A˚ absorption is a reliable youth indi- We noticed that a significant fraction of stars with es- cator in stars later than G (e.g. Duncan & Jones 1983; timated ages higher than average for the region have Favata et al. 1998). We defer from using Li EW to esti- masses of 1 M . This leads us to suggest that the age mate age, as the effect of lower metallicity in the spectral spread could∼ possibly be due to errors in birthline cal- range under study on the lithium boundary is unclear. culations for G-type stars (Hartmann 2003), which may Errors in spectral type determination larger than two- lead to a systematic overestimation of absolute stellar three sub-classes for these stars could lead to incor- ages. Therefore, any ages derived can only be accurate rect temperature determinations, but also extinction and in a relative sense. Any such birthline effect is circum- bolometric corrections, and thereby luminosity. How- vented by interpolating the ages of solar metallicity PMS ever, the two effects roughly cancel leading to small ab- stars used for comparison using the same isochrones and solute age errors (see Hartmann 2001). Moreover, any bi- following the same interpolation technique. This is de- nary companions would roughly double the measured lu- scribed further in Appendix A. minosity, while keeping the temperature the same, leading to younger ages. Another source of error may in- 3.2. Accretion luminosity volve the stellar isochrones themselves. Differences in input physics imply that absolute ages derived using dif- In the current PMS evolutionary paradigm, stars ac- ferent isochrones could differ by as much as a factor of 2. crete mass from a surrounding circumstellar disk. The We tested this by comparing the ages derived using the disk is thought to be curtailed at some inner radius (Rin) Bressan et al. (2012) isochrones from the PISA database by the stellar magnetosphere, and gas in the disk is ac- creted via the magnetic field lines. The gravitational en- 6 V. M. Kalari & J. S. Vink

6.5 ergy is released, as material falls onto the stellar surface − creating an accretion shock, leading to excess continuum emission, which is particularly evident in the ultraviolet wavelength range. The ionization and recombination of 7.0 gas during this process leads to strong Balmer line emis- − ) 1 sion (notably in Hα). In this scenario, the bolometric − yr

accretion luminosity (Lacc) can be estimated from the M

re-radiated energy, which can be measured from the Hα ( 7.5 − line luminosity (LHα). acc ˙ We measure LHα by adapting the method of Mohanty M et al. (2005) (see also Costigan et al. (2014) for an ap- log U plication to higher-mass stars). The flux from Castelli & 8.0 Kurucz (2004) model spectra (Z = 0.004) for the nearest − point in (Teff ,log g) plane in 6554 - 6572 A˚ range is taken as continuum flux. We adopt for all stars log g = 4.0. The continuum flux is multiplied by the Hα EW and 8.5 2 − 8.5 8.0 7.5 7.0 6.5 4πR to obtain LHα. Veiling is estimated by compar- − − − 1 − − Hα log M˙ acc (M yr− ) ing with Pickles (1998) non-accreting stellar spectra. In the range with measurable absorption lines (> 5800 A),˚ Fig. 5.— Comparison of M˙ acc measured from Hα spectroscopy veiling is negligible. This is in good agreement with the (abscissa) with archival U-band photometry (ordinate). The average errors are shown in the top-left hand corner. Although the results of Calvet et al. (2004) for stars in a similar wave- U-band sample is small, the comparison suggests consistency be- length and mass range, but using high-resolution spectra tween mass-accretion rates determined from Hα and the U-band, over a much larger wavelength range. We adopt the em- i.e. Hα is most likely a good indicator for mass-accretion phenom- ena. pirical LHα-Lacc relation used by De Marchi et al. (2010);

log Lacc = log LHα + 1.72 0.47 (1) Five stars out of the total 24 have archive U-band pho- ± tometry in Delgado et al. (2010). The excess U-band The relation was derived by comparing Lacc estimated luminosity (LU,excess) of these stars, thought to be due from the UV continuum-excess with LHα of 0.5 - 2 M to accretion is estimated according to the formula mass stars (Dahm 2008). log Lacc estimated this way in- 2 2 cludes an uncertainty of factor 0.5, which is the dominant LU,excess = 4πW(d fobs R fmod) (3) − ∗ error in the measured M˙ acc. The LHα-Lacc used by De (Romaniello et al. 2002). fmod and fobs are the model Marchi et al. (2010) agrees within errors with that in a and observed fluxes in the U-band respectively. W is the similar range derived by Mendigut´ıaet al. (2011). We width of the U-band filter. The LU,excess is translated stress that the LHα-Lacc relation was adopted with a view into L following the relation of Gullbring et al. (1998) to reduce any sources of systematic errors between our acc results to those in the LMC. A comparison of the method log Lacc = 1.09 log LU,excess + 0.98. (4) used to calculate L with respect to traditional Hα line acc ˙ flux measurements is presented in Appendix B. Using eq. (2), we compute Macc. A comparison of the M˙ acc derived using U-band photometry and Hα line lu- 3.3. Mass accretion rate minosity are presented in Fig. 5. We find that the dif- M˙ can be estimated from the L using the free-fall ference is not greater than the mean error (0.5 dex), sug- acc acc ˙ equation; gesting that the Macc is reasonably well constrained for our purposes.

LaccR Rin M˙ acc = ∗ . (2) 3.4. Disk properties GM Rin R ! ∗ − ∗ We have shown that all 24 PMS stars in Sh 2-284 have Here M , R are the stellar mass and radius respec- Hα in emission, and for a subset the accretion rates de- ∗ ∗ tively. The inner radius (Rin) is uncertain and depends rived from Hα agree with those derived from the U-band on the coupling of the accretion disk with the magnetic (see also Kalari et al. 2014, in prep.). This demonstrates ﬁeld lines. We follow Gullbring et al. (1998) and take to a reasonable degree that these stars are accreting. Rin = 5 2 R (Vink et al. 2005). Using the stellar mass Most of these stars also have near-infrared 2MASS JHKs and radius± measured from the H-R diagram, we estimate magnitudes (Cutri et al. 2003), and Spitzer imaging up to 8µm (Puga et al. 2009). In this section, we analyze the M˙ acc for each star. The accretion properties are tab- ulated in Table 3. the disk properties of the PMS objects using infrared photometry. As a sanity check on the estimated M˙ acc, we compared whenever possible with archival U-band photome- The slope of the SED in the infrared, ˙ try of Delgado et al. (2010). Macc, or more directly the d log(λF ) α = λ , (5) Lacc can be measured reliably from U-band photometry. IR d logλ Comparison of the observed U-band magnitude with a template magnitude comprises a direct measure of the at λ > 3 µm is used to diagnose the evolutionary stage excess U-luminosity caused by the emission shock which of the disk-star system (Lada 1987). We adopt the clas- can be translated into Lacc (Gullbring et al. 1998). siﬁcation scheme of Greene et al. (1994) to distinguish Pre-main sequence accretion rates at low metallicity 7

TABLE 3 Accretion and disc properties of PMS stars in Dolidze 25

No. ID (2MASS) Mass Age log Lacc log M˙ acc αIR Class 1 1 (M ) (Myr) (L yr− )(M yr− )

1 J06443291+0023546 1.79 4.0 0.28 6.92 0.27 II 2 J06443294+0010528 1.83 3.5 0.95 −6.24 −0.28 II 3 J06443541+0019093 1.57 5.6 0.71 −7.93 −2.02 III 4 J06443682+0016186 1.6 5.5 −0.26 − 7.5 −0.89 II 5 J06443788+0021509 1.79 3.9 −0.65 −7.85− - III 6 J06443827+0019229 1.26 10.9 − 0.6 − 7.9 1.41 II 7 J06444496+0019335 1.75 3.3 −0.46 −7.69 −0.97 II 8 J06444602+0019182 1.57 3.8− 0.29 −6.96− - III 9 J06444714+0013320 0.91 0.9 0.6 −7.61 0.76 II 10 J06444936+0020245 1.36 4.3 −0.67 −7.95− 0.59 I 11 J06445577+0013168 2.08 2.7 −0.04 − 7.2 0.38 I 12 J06445837+0014151 2.0 3.0− 0.38 −6.79 0.70 I 13 J06450075+0013356 1.05 1.5 0.49 −7.61 1.25 II 14 J06450208+0019443 1.75 4.1− 0.48 −6.82− 0.92 I 15 J06450274+0018077 1.3 9.5 0.07 −7.36 0.23 II 16 J06450681+0013535 2.16 2.0 −0.92 −8.07 − 0.3 II 17 J06450971+0014121 2.62 1.5− 0.53 −6.66− - II 18 J06451318+0018307 1.15 9.3 0.38 − 7.7 1.09 II 19 J06451616+0022238 1.23 14.8 −1.09 −8.46− - II 20 J06451701+0022077 2.0 7.5 −0.07 −7.45 - II 21 J06451727+0023454 1.57 5.6 −0.11 −7.33 0.32 II 22 J06451841+0022189 1.12 8.5 −1.36 −8.68 −0.52 II 23 J06452454+0022448 1.66 1.0 −0.18 −7.35 −0.98 II 24 J06452774+0015514 0.56 0.9 − 0.8 −7.66 −0.77 II − − − between systems with protostellar disks (Class I), opti- properties of Z Galactic and 1/3Z LMC PMS stars to cally thick disks (Class II), or no circumstellar excess examine if M˙ acc estimates are a function of Z within the (Class III). αIR was derived by fitting the Spitzer magni- observed Z range. tudes at 3.6, 4.5, 5.8, and 8µm. The resultant αIR values are given in Table 3. The disk SEDs are shown in Fig. 6. 4.1. Lacc vs. L Five stars do not have photometry in any band. For ∗ these stars, we fitted the available Spitzer magnitudes to The detection limit on M˙ acc are dependent on the de- the disk models of Robitalle et al. (2006) to classify the tection limits of the measured Lacc. This is demonstrated disk evolutionary state. in the Lacc vs. L distribution in Fig. 7. We find that ∗ Out of the 19 accreting PMS stars that have Spitzer Lacc < L , i.e., we did not identify any continuum stars photometry at all observed wavelengths, we find that 18 (Hernándezet∗ al. 2004) in our sample, leading to an have SED slopes resembling Class I/II objects, with one upper boundary in the M˙ acc-M plane, possibly leading star having αIR resembling a Class III object. Three ∗ to a scarcity of high M˙ acc. However, it is thought that stars do not have 8 µm photometry (J06443788+0021509, continuum stars form only a very small fraction of the J06444602+0019182), of which two have best-fit disk total population in well-studied regions such as Taurus models suggestive of either flat or sources devoid of any (Hartmann 2008). infrared excess, and one (J06451701+0022077) sugges- The lower detection limits have been calculated from tive of a dusty disk. Two stars that do not have pho- L and M of stars lying on the 3 Myr isochrone with tometry at either 4.5, or 5.8 µm (J06451701+0022238, T ∗ = 5000,∗ 6000, 7000, 8000, 10000 and 12000 K. We J06450971+0014121), have best-fit models suggestive of eff assume a Hα EW = 2 A˚ to calculate their accretion lu- Class II objects. Overall, most of the PMS star sample minosity. The limit− is near the point where accretion have infrared-excess akin to circumstellar disks, verifying is not detectable in Hα line emission. We note that at that we deal with genuine accreting PMS stars. Lacc lower than this boundary it would be impossible to detect accretion using conventional means such as line 4. DISCUSSION emission or UV-excess, since Hα is the most sensitive in- We present spectroscopic estimates of accretion rates dicator to tiny accretion rates. The lower boundary of of 24 low-Z PMS stars in a stellar mass range between Lacc is within the error bars of detections at L < 1 L , 1 - 2 M , having a median age of 3.5 Myr, and a median at approximately 1.5 M . The PMS stars identified∗ by 7.5 1 accretion rate of 10− M yr− . The majority of our Cusano et al. (2011) which form the bulk of our sam- sample (21 out of 24 stars) shows evidence for circum- ple stars, were identified by the authors of that paper stellar disks, and all stars with literature U-band pho- as those stars showing Hα emission out of 1506 spec- tometry (5 out of 24) display an UV excess, confirming tra observed in the region. Stars were rejected as non- the PMS nature of the sample. Based on our results, members based on their luminosity classification, and in- we discuss in the following section the relation between frared colors. It is plausible to expect that the Cusano et Lacc and stellar luminosity; and the observed dependence al.(2011) PMS sample would contain stars having lower of the estimated M˙ acc on stellar mass and age. We also Lacc at higher masses if they existed. This suggests that compare our results with literature estimates of accretion our sample contains PMS stars at the lowest measurable 8 V. M. Kalari & J. S. Vink

110.0 10 − − 11 1 2 − 3 11 11 − − 12 − 12 12 13 − − − 0.3 0.1 0.5 0.9 0.3 0.1 0.5 0.9 0.3 0.1 0.5 0.9 − − − 10 10.5 − − 10.5 4 − 5 6 11 11.5 − 11.5 − − 012.8 12.5 12.5 − − − 0.3 0.1 0.5 0.9 0.3 0.1 0.5 0.9 0.3 0.1 0.5 0.9 − − − 10.5 11.5 11 − − − 7 8 9 11.5 − 12.5 12 − − 12.5 − 13.5 13 0.3 0.1 0.5 0.9 − 0.3 0.1 0.5 0.9 − 0.3 0.1 0.5 0.9 − − − 11 10 − − 0.6 9.5 ) 10 − 11 12 1

− 11

s 12 10.5 −

2 − − − 11.5 12 − − 13 − 0.3 0.1 0.5 0.9 0.3 0.1 0.5 0.9 0.3 0.1 0.5 0.9 − − − 12 10 10.5 (ergs cm − − −

λ 13 14 15 11 11.5 λ F 13 − − −

log 12 12.5 0.4 − − 14 − 0.3 0.1 0.5 0.9 0.3 0.1 0.5 0.9 0.3 0.1 0.5 0.9 − − − 10 11 − − 10 16 − 17 18 11 12 − 11 − − 12 12 13 − − − 0.3 0.1 0.5 0.9 0.3 0.1 0.5 0.9 0.3 0.1 0.5 0.9 − − − 10 − 10.5 10 − 0.2 19 20 − 21 11 11.5 − 11 − − 12.5 12 − − 12 0.3 0.1 0.5 0.9 0.3 0.1 0.5 0.9 − 0.3 0.1 0.5 0.9 − − − 11.5 11 11.5 − − − 22 23 24 12.5 12 12.5 − − −

130.50 13 13.5 − 0.0 0.3 0.1 0.5 0.2 0.9 − 0.43 0.1 0.5 0.09.6 − 0.3 0.80.1 0.5 0.9 1.0 − − − log λ (µm)

Fig. 6.— The spectral energy distribution of Sh 2-284 pre-main sequence stars. The number in the top-right corner corresponds to the index number in Table 3. Dashed lines are the Kurucz & Castelli (2004) model spectra corresponding to the determined spectral parameters in Table 2. Dots are the un-reddened U,B,V,R,I fluxes from Delgado et al. (2010) or Cusano et al. (2011). Crosses are JHKs fluxes from 2MASS (Cutri et al. 2003) and Spitzer IRAC fluxes at 3.6, 4.5, 5.8, and 8.0 µm. Pre-main sequence accretion rates at low metallicity 9 0.3 Myr 1 Myr

tic PMS stars is shown in Fig.3 Myr8. The Sh 2-284 PMS stars occupy similar ranges3 M to ZZAMSPMS stars in both 1.5 ˙ ˙ 2 M the Macc-M and Macc-t plane. We find an absence of ∗ 1.5∗ M 1.0 LMC PMS stars at low M˙ acc which suggest that the over- 0.8 M all slopes measured by Spezzi et al. (2012) may be sig- 0.5 ) nificantly affected by detection limitations. This could L ( be due to the difficulty of observing low-mass stars in 0.0 10 Myr acc the distant LMC. We also find an absence of stars with L ˙ 7 1

log 0.5 Macc > 10− M yr− at ages beyond 5 Myr in our sample, a region of the plane that is populated by points 1.0 from the LMC.

1.5 4.3. Is M˙ acc a function of Z?

2.0 ˙ 0.0 0.5 1.0 1.5 2.0 We have shown that the Macc in Sh 2-284 and Galactic log L (L ) solar-Z star forming regions follow similar distributions in the M˙ acc-M and M˙ acc-t plane. The PMS star sample Fig. 7.— Lacc as function of L . The dashed line displays the ∗ ∗ ∗ in the LMC of Spezzi et al. (2012) have shallower slopes, L = Lacc relation. The inverted triangles give the detection limits ∗ which may be affected by detection limits. However, as calculated for six different Teff for model L at 3 Myr. The upper ∗ left corner displays the mean error bars. The Lacc are given in discussed by Spezzi et al. (2012), the median M˙ acc at any Table 3. given age in a defined mass range (1-2 M ) is a robust ˙ Macc within this range. One star lies at the boundary of tool for comparing any differences in M˙ acc between star- the lower detection limit (J06450681+0013535) and has forming regions. To investigate this we tabulate the me- ˚ a measured Hα EW of 1.5 A, lower than the assumed dian M˙ acc of 1 - 2 M PMS stars in known star-forming threshold. − regions in Table 4. The median age for the given sample is computed from the isochronal ages. To estimate the ˙ 4.2. Macc as a function of stellar mass and age completeness of each sample, we calculate the expected fraction of stars within this mass range using the Kroupa M˙ acc are plotted as a function of M in Fig. 8. We find the best-fit power law relationship across∗ the 1 - 2 M (2001) IMF, and total cluster masses from Weidner & 2.4 0.35 Kroupa (2006). The median age of each star-forming range to be M˙ acc M ± . There is a spread of ∝ ∗ ∼ region is used to determine the expected fraction of ac- 2 dex in M˙ acc at any given M . The spread can partially ∗ cretion stars, Nexp (Fedele et al. 2010). be related to the intrinsic distribution of ages at any ˙ ˙ The median log Macc of Sh 2-284 PMS stars is given mass, as Macc vary with age. But the scatter is also 7.5 dex. It is slightly larger than the Galactic mass- found in coeval populations, suggesting that there are accretion− rate by 0.3 dex in the IC 348, Trumpler 37 other factors beyond mass and age controlling accretion region and Cha II∼ clouds, but similar to the values ob- rates (Hartmann et al. 2006). tained in the Taurus-Auriga region and Lupus molecular We employed survival analysis linear regression using ˙ ASURV (Lavalley et al. 1992) to include lower and clouds. In comparison, the median log Macc reported by higher limits in accretion rates. We find that includ- Spezzi et al. (2012) in an LMC PMS population having a ing lower limits leads to a steeper slope, but within median age of 6 Myr is 7.1 dex. Using Equation 7 from − ˙ the error bar. The inclusion of upper limits leads to De Marchi et al. (2013), the median log Macc expected in a shallower slope, but within 0.1 dex of the slope calcu- the SMC cluster NGC 602 at an age of 3 Myr for 1 - 2 M is expected to be 7.1 dex. lated. This suggests that the absence of low-mass high- & − accretion rates, or high-mass low-accretion rates is not Based on Table 4 we see that the median M˙ acc in Sh statistically significant in our regression fitting, and that 2-284 are comparable to Galactic values at similar ages. the given distribution statistically represents a general There is no systematic difference at an age of 3 Myr in trend of increasing M˙ acc with M . The index α mea- the accretion rates of solar Z or Sh 2-284 PMS stars sured is close to the Galactic median∗ 2, but deviant based on these results. If we assume that our results are from the value of 1 claimed in metal-poor∼ PMS stars not significantly underestimated (see Section 4.2), it is in the LMC (Spezzi∼ et al. 2012). challenging to explain physically how accretion is pro- The ages of stars in the Sh 2-284 region are around 3 ceeding at much higher rates in much older stars in the Myr in the central cluster, but vary in the surrounding LMC, especially considering that ZLMC 2 ZSh 2 284. ∼ × − nebula (Puga et al. 2009). The age distribution of Sh It seems unlikely that Z is the only quantity determin- ˙ ing the accretion rates, and factors other than Z may be 2-284 is reflected in the Macc-Age plane (Fig. 8). We fit ˙ the orthogonal distance regression to account for errors responsible for the high Macc measured by the studies of De Marchi et al. (2010; 2011a;2011b;2013) and Spezzi et on both M˙ acc and t . We find the power-law index η to be 0.7 0.4, similar∗ to the slope found in studies al. (2012). − ± There are practical considerations which might lead to of Z PMS stars, and as expected in line with viscous ˙ disk evolution at these masses (Sicilia-Aguila et al. 2006; higher measured Macc at low Z. The observable limit ˙ Manara et al. 2012). for the highest measurable Macc is generally given by Overall, a comparison of Sh 2-284 PMS stars with the Lacc < L , owing to the difficulty in detecting continuum ∗ literature accretion rates of 1/3 Z LMC and Z Galac- stars, and the lowest M˙ acc at Hα EW < 1 A˚ beyond which

10 V. M. Kalari & J. S. Vink

6.0 6.0 − −

6.5 6.5 − −

7.0 ) ) 1 7.0 1 − − − − yr yr

7.5 M

M − ( 7.5 ( − acc acc ˙ ˙ 8.0 M M − 8.0 log − log 8.5 − 8.5 − 9.0 − 1 M < M < 2 M 9.0 ∗ − 9.5 0.05 0.00 0.05 0.10 0.15 0.20 0.25 0.30 − 5.5 6.0 6.5 7.0 7.5 − log M (M ) log t (yr) ∗

Fig. 8.— Distribution of M˙ acc as function of M (left) and t (right) shown as blue circles. The solid line is the best-ﬁt regression slope ∗ ∗ in the 1 - 2 M range, and has an index α = 2.4 0.35 in the M˙ acc-M plane and index η = 0.7 0.4 in the M˙ acc-t plane. The dashed line (right) is the expected evolution of viscous disks± following Hartmann∗ et al. (1998), and left is± the slope for LMC∗ PMS stars reported in Spezzi et al. (2012). Black inverted triangles are the detection limits. Triangles are the data points for Galactic PMS taken from the studies of Natta et al. (2006); Herczeg & Hillenbrand (2008); Sicilia-Aguilar et al. (2010); Barentsen et al. (2011; 2013); Antoniucci et al. (2011); Manara et al. (2012); Alc´aleet al. (2014). Red crosses are taken from the studies of LMC PMS stars from Spezzi et al. (2012).

TABLE 4 Median M˙ acc of PMS stars in the 1 -2 M at different Z

Region N Nexp Age log M˙ acc Reference 1 (Myr) (M yr− )

Z = Z

L1641 23 - 1 7.1 Fang et al. (2009); − Caratti o Garatti et al. (2012) Orion nebula cluster 41 75 10 1.5 7.65 Manara et al.(2012) ρ-Ophiuchus 5 5±3 0.5-2a −7.95 Natta et al.(2006) NGC 2264 7 15± 4 3 − 7.8 Barentsen et al.(2013) Taurus-Auriga 3 2 ±1 3 −7.4 Calvet et al.(2004) IC 348 2 6±2 1.5-4b −7.8 Dahm (2008) Lupus 3± - 3 −7.6 Alcal´aet al.(2014); − Herczeg & Hillenbrand (2008) Cha II 7 - 3-4 7.9 Antoniucci et al.(2011) Trumpler 37 42 (b) 4 −7.85 Sicilia-Aguilar et al.(2010); − Barentsen etal.(2011) Z = 0.4 Z

LMC Field3 Pop.2 146 6 7.2 Spezzi et al.(2012) LMC Field3 Pop.1 96 8 −7.2 Spezzi et al.(2012) LMC Field2 325 9 −7.4 Spezzi et al.(2012) LMC Field1 490 13 −7.6 Spezzi et al.(2012) SN 1987 A 104 14 −7.6 De Marchi et al.(2010) − Z = 0.2 Z

Sh 2-284 20 3.5 7.5 This work − Z = 0.1 Z

NGC 346d - 1 7.1 De Marchi et al.(2011) NGC 346d - 20 −7.7 De Marchi et al.(2011) −

Note. — (a) ρ-Ophiuchus is oft-quoted with an age of 0.5 Myr (Natta et al. 2006), however the ages of the accreting PMS stars are thought to be as old as 2 Myr (Wiliking et al. 2008; Rigiliaco et al. 2010). (b) The total cluster mass of Trumpler 37 is debated, as studies of stellar content reveal a dearth of near solar mass members, and a deviant stellar mass function slope (Errman et al. 2012). (c) The age of IC 348 is debated, with initial studies suggesting an age as young as 1.5 Myr (Herbig 1998), but later revised to around 4 Myr (Mayne & Naylor 2007). (d) Taken from Figure 8 of Spezzi et al. (2012). Pre-main sequence accretion rates at low metallicity 11 accretion is hard to measure. The lower limit leads to a If accretion if present beyond such rates it is not eas- threshold in L (shown in Fig. 7). L increases inversely ily detectable using current methods. We find M˙ acc in a non-monotonic∗ fashion as a function∗ of Z according to steeply increase with M , with a power-law index to stellar isochrones. This suggests that we are measur- 2.4 0.35 when considering∗ only stars in the 1 - 2 M ing stars of similar spectral type, but with different L at ± ˙ ∗ range. While Macc decrease with stellar age, with an different Z, with consequently higher Lacc and thereby index 0.7 0.4. There exists a spread 2 orders of ˙ Macc at lower Z. This also leads to raising the Lacc - L magnitude− in± the distribution of sources, which∼ cannot ∗ upper limit shown in Fig. 7, and leading to a slightly be singularly attributed to any one effect. higher M˙ acc. To test the difference that may be caused Through comparison with solar-Z PMS stars, we find by this effect, we calculate the lower detection limit us- no significant differences in the 1 - 2 M range. The dis- ing solar Z isochrones and use these to measure the tribution of sources in the M˙ acc-M and M˙ acc-t planes ˙ ∗ ∗ median Macc for censored data using a Kapalan-Meier are very similar, although there are fewer lower M˙ acc ob- test. We find that this leads to a slightly lower median jects at younger ages in Sh 2-284. However, this may be M˙ acc 7.7 dex, suggesting that there is a non-negligible a product of low number statistics at very young ages. ∼ − difference in detection limits of PMS stars at different Z. The median log M˙ acc of Sh 2-284 ( 7.5 dex) is in agree- This work is based on the assumption that the metallic- − ment with the median M˙ acc of 3 Myr old solar-Z star ity of the PMS stars in Sh2-284 is similar to that of the B forming regions. When compared to the results in the stars located in the central open cluster Dolidze 25. The LMC (Z = 0.008) from Spezzi et al. (2012), we find that metallicity of Dolidze 25 is anomalous when taken into ˙ account the distance-metallicity gradient of the Galaxy the median log Macc of LMC PMS stars are 7.2 dex at ˙ ∼ − (Rolleston et al. 2000). It has been suggested that the 6 Myr, suggesting that the median Macc are much lower close association with the Canis Major Dwarf Galaxy is in Sh 2-284. We also find that the distribution of Sh 2-284 sources in in the M˙ acc-M and M˙ acc-t planes occupy re- the reason for the metal-deficient nature of the central ∗ ∗ OB stars (Martin et al. 2004). It seems reasonable to gions at lower M˙ acc when compared to LMC stars. This assume that the metallicity of the PMS stars in the Sh2- difference is not easily explained, particularly when we 284 region is similar to the central B stars, however, if consider that the Z of the LMC is twice that of Sh 2-284, other processes lead to the metal-deficient nature of the and if Z is truly proportional to M˙ acc, M˙ acc of Sh 2-284 central B stars, the PMS stars may have higher abun- PMS stars would be similarly higher than the LMC aver- dances. Future work on the Z determination of PMS age, as the LMC measures are compared to the Galaxy. stars in Sh2-284 would thus be welcomed. It is likely that the Spezzi et al. (2012) results are influenced by completeness issues. Moreover, it is not clear 5. SUMMARY AND FUTURE WORK whether accretion can be sustained at such high rates for We presented an optical spectroscopic study of PMS such long periods as measured in the LMC PMS stars, stars located in the star-forming region Sh 2-284. Sh which, if true would suggest different disk evolutionary 2-284 is a Galactic star-forming region with Z 0.004, mechanisms in comparison to Galactic PMS stars, where i.e. 1/5 Z . Our sample consists of 24 objects∼ span- accretion is thought to cease by around 5 Myr. We sug- ning a mass range between 0.9 - 2.6 M and an age range gest that there are most likely factors beyond differences between 1 - 10 Myr, with a median age of 3.5 Myr. Our in Z which influence the measured differences in M˙ acc. results are consistent with the scenario of sequential star These include practical differences in detection limits at formation suggested by Puga et al. (2009) inferred from different Z, and difficulty in identifying PMS stars at gas dynamics. 21 stars have significant infrared excess distances to the LMC. A more systematic and thorough up to 8µm, suggesting that they are bonafide Class I/II study of PMS stars in the Magellanic Clouds and the PMS stars, whilst the remaining PMS stars are most Galaxy is required to disentangle the various factors that likely Class III objects. influence measured mass-accretion rates. We used the measured Hα EW, and stellar parameters to derive LHα, from which we estimated the M˙ acc. For five objects, we also have archival U-band photometry, Based on observations obtained with the South African and we measured the U-band excess intensity from which Large Telescope under ID 2013-2-UKSC-008; and the we estimated M˙ acc. We found that the M˙ acc estimated ESO telescopes at the La Silla Paranal Observatory un- from the LHα agree well with those derived from the U- der program no. 0.74.C-0111. We would like to thank the band excess. The multiple signatures of disk accretion referee and Drs. Antonella Natta and Danny Lennon for are highly suggestive of ongoing accretion in the Sh 2- useful comments. V.M.K. is the recipient of a postgradu- 284 PMS stars. ate scholarship awarded by the Department for Culture, Our sample is detection limited in Lacc and M˙ acc. The Arts and Leisure Northern Ireland. J.S.V. and V.M.K. highest measured Lacc is never greater than L , and the acknowledge funding from the the Science and Technol- lowest measured L corresponds to a Hα EW∗ 1 A.˚ ogy Facilities Council U.K. acc ∼ − REFERENCES Alcalá,J. M., Natta, A., Manara, C. F., et al. 2014, A&A, 561, Barentsen, G., Vink, J. S., Drew, J. E., & Sale, S. E. 2013, AA2 MNRAS, 429, 1981 Antoniucci, S., Garc´ıaLópez, R., Nisini, B., et al. 2011, A&A, Barentsen, G., Vink, J. S., Drew, J. E., et al. 2011, MNRAS, 415, 534, AA32 103 12 V. M. Kalari & J. S. Vink

Beaulieu, J. P., Lamers, H. J. G. L. M., Grison, P., et al. 1996, Kalari, V. M., Vink, J. S., Dufton, P. L., et al. 2014, A&A, 564, Science, 272, 995 L7 Bressan, A., Marigo, P., Girardi, L., et al. 2012, MNRAS, 427, 127 Kenyon, S. J., & Hartmann, L. 1995, ApJS, 101, 117 Burgh, E. B., Nordsieck, K. H., Kobulnicky, H. A., et al. 2003, Kroupa, P. 2001, MNRAS, 322, 231 Proc. SPIE, 4841, 1463 Lada, C. J. 1987, Star Forming Regions, 115, 1 Calvet, N., Muzerolle, J., Briceño,C., et al. 2004, AJ, 128, 1294 Lamers, H. J. G. L. M., Beaulieu, J. P., & de Wit, W. J. 1999, Cai, K., Durisen, R. H., Michael, S., et al. 2006, ApJ, 636, L149 A&A, 341, 827 Caratti o Garatti, A., Garcia Lopez, R., Antoniucci, S., et al. Lavalley, M., Isobe, T., & Feigelson, E. 1992, Astronomical Data 2012, A&A, 538, AA64 Analysis Software and Systems I, 25, 245 Castelli, F., & Kurucz, R. L. 2004, arXiv:astro-ph/0405087 Lennon, D. J., Dufton, P. L., Fitzsimmons, A., Gehren, T., & Costigan, G., Vink, J. S., Scholz, A., Ray, T., & Testi, L. 2014, Nissen, P. E. 1990, A&A, 240, 349 MNRAS, 440, 3444 Manara, C. F., Robberto, M., Da Rio, N., et al. 2012, ApJ, 755, Cusano, F., Ripepi, V., Alcalá,J. M., et al. 2011, MNRAS, 410, 154 227 Martin, N. F., Ibata, R. A., Bellazzini, M., et al. 2004, MNRAS, Cutri, R. M., Skrutskie, M. F., van Dyk, S., et al. 2003, VizieR 348, 12 Online Data Catalog, 2246, 0 Mendigut´ıa,I., Calvet, N., Montesinos, B., et al. 2011, A&A, 535, Dahm, S. E. 2008, AJ, 136, 521 A99 De Marchi, G., Beccari, G., & Panagia, N. 2013, ApJ, 775, 68 Mohanty, S., Basri, G., & Jayawardhana, R. 2005, Astronomische De Marchi, G., Paresce, F., Panagia, N., et al. 2011a, ApJ, 739, 27 Nachrichten, 326, 891 De Marchi, G., Panagia, N., Romaniello, M., et al. 2011b, ApJ, Monet, D. G., Levine, S. E., Canzian, B., et al. 2003, AJ, 125, 984 740, 11 Muzerolle, J., Luhman, K. L., Briceño,C., Hartmann, L., & De Marchi, G., Panagia, N., & Romaniello, M. 2010, ApJ, 715, 1 Calvet, N. 2005, ApJ, 625, 906 de Wit, W. J., Beaulieu, J. P., Lamers, H. J. G. L. M., Coutures, Natta, A., Testi, L., & Randich, S. 2006, A&A, 452, 245 C., & Meeus, G. 2005, A&A, 432, 619 Parker, Q. A., Phillipps, S., Pierce, M. J., et al. 2005, MNRAS, de Wit, W. J., Beaulieu, J.-P., Lamers, H. J. G. L. M., Lesquoy, 362, 689 E., & Marquette, J.-B. 2003, A&A, 410, 199 Palla, F., & Stahler, S. W. 1999, ApJ, 525, 772 de Wit, W. J., Beaulieu, J. P., & Lamers, H. J. G. L. M. 2002, Palla, F., & Stahler, S. W. 1990, ApJ, 360, L47 A&A, 395, 829 Panagia, N., Romaniello, M., Scuderi, S., & Kirshner, R. P. 2000, Delgado, A. J., Djupvik, A. A., & Alfaro, E. J. 2010, A&A, 509, ApJ, 539, 197 A104 Pickles, A. J. 1998, PASP, 110, 863 Duncan, D. K., & Jones, B. F. 1983, ApJ, 271, 663 Pogodin, M. A., Hubrig, S., Yudin, R. V., et al. 2012, Ercolano, B., & Clarke, C. J. 2010, MNRAS, 402, 2735 Astronomische Nachrichten, 333, 594 Fang, M., van Boekel, R., Wang, W., et al. 2009, A&A, 504, 461 Pringle, J. E. 1981, ARA&A, 19, 137 Favata, F., Micela, G., Sciortino, S., & D’Antona, F. 1998, A&A, Puga, E., Hony, S., Neiner, C., et al. 2009, A&A, 503, 107 335, 218 Robitaille, T. P., Whitney, B. A., Indebetouw, R., Wood, K., & Fedele, D., van den Ancker, M. E., Henning, T., Jayawardhana, Denzmore, P. 2006, ApJS, 167, 256 R., & Oliveira, J. M. 2010, A&A, 510, A72 Rolleston, W. R. J., Smartt, S. J., Dufton, P. L., & Ryans, Fitzsimmons, A., Dufton, P. L., & Rolleston, W. R. J. 1992, R. S. I. 2000, A&A, 363, 537 MNRAS, 259, 489 Romaniello, M., Robberto, M., & Panagia, N. 2004, ApJ, 608, 220 Greene, T. P., Wilking, B. A., Andre, P., Young, E. T., & Lada, Romaniello, M., Panagia, N., Scuderi, S., & Kirshner, R. P. 2002, C. J. 1994, ApJ, 434, 614 AJ, 123, 915 Gullbring, E., Hartmann, L., Briceño,C., & Calvet, N. 1998, Sharpless, S. 1959, ApJS, 4, 257 ApJ, 492, 323 Sicilia-Aguilar, A., Henning, T., & Hartmann, L. W. 2010, ApJ, Hartmann, L. 2008, Accretion Processes in Star Formation, by 710, 597 Lee Hartmann, Cambridge, UK: Cambridge University Press, Sicilia-Aguilar, A., Hartmann, L. W., Fürész,G., et al. 2006, AJ, 2008, 132, 2135 Hartmann, L., D’Alessio, P., Calvet, N., & Muzerolle, J. 2006, Spezzi, L., de Marchi, G., Panagia, N., Sicilia-Aguilar, A., & ApJ, 648, 484 Ercolano, B. 2012, MNRAS, 421, 78 Hartmann, L. 2003, ApJ, 585, 398 Stahler, S. W. 1983, ApJ, 274, 822 Hartmann, L. 2001, AJ, 121, 1030 Tognelli, E., Prada Moroni, P. G., & Degl’Innocenti, S. 2011, Hartmann, L., Calvet, N., Gullbring, E., & D’Alessio, P. 1998, A&A, 533, A109 ApJ, 495, 385 Tout, C. A., Livio, M., & Bonnell, I. A. 1999, MNRAS, 310, 360 Hartmann, L., Cassen, P., & Kenyon, S. J. 1997, ApJ, 475, 770 Vink, J. S., Drew, J. E., Harries, T. J., Oudmaijer, R. D., & Herczeg, G. J., & Hillenbrand, L. A. 2008, ApJ, 681, 594 Unruh, Y. 2005, MNRAS, 359, 1049 Hernández,J., Calvet, N., Briceño,C., Hartmann, L., & Berlind, Weidner, C., & Kroupa, P. 2006, MNRAS, 365, 1333 P. 2004, AJ, 127, 1682 Yasui, C., Kobayashi, N., Tokunaga, A. T., Saito, M., & Tokoku, C. 2009, ApJ, 705, 54

The Lacc quoted by the Spezzi et al. (2012) and Manara et al. (2012) study were combined with the appropriate APPENDIX M to derive M˙ using Equation 2. In Figs. 9 and 10, ˙ acc AGES AND MACC DERIVED USING DIFFERENT ∗ ˙ STELLAR TRACKS AND ISOCHRONES we compare the Macc and ages derived using the Bres- san et al. (2012) stellar models with those quoted in the We use the PMS isochrones and tracks from Bressan studies. Namely, Spezzi et al. (2012) employed the Pisa et al. (2012) to derive the ages and masses for our sam- isochrones database (Tognelli et al. 2011); while Manara ple. To account for differences in stellar models when et al. (2012) derived results using a variety of isochrones, comparing our results with literature results, we esti- where we chose the results they derived from employing mate the M˙ acc, M , and t using the same Bressan et the Palla & Stahler isochrones (1999). We find small sys- ∗ ∗ al. (2012) stellar tracks and isochrones at the appropri- tematic offsets in the ages and M˙ acc when compared to ate Z. We used the quoted log L, and Teff of PMS stars the results of Spezzi et al. (2012), but no systematic dif- in the LMC studies of Spezzi et al. (2012), and the Orion ferences in M˙ acc compared to Manara et al. (2012). We Nebula study by Manara et al. (2012) to estimate M find a non-systematic change in the ages of Orion Nebula and t using the Bressan et al. isochrones and tracks.∗ ∗ Pre-main sequence accretion rates at low metallicity 13

8.0 9.0 −

8.5 7.5 −

8.0 − 7.0 7.5 − 6.5 7.0 −

) from Spezzi et6 al.. 2012 5 6.0 1 − − yr

6.0

M − 5.5 ( acc

˙ 5.5 log Age (yr) from Spezzi et al. 2012 M −

5.0 log 5.0 −

4.5 4.5 − 4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0 8.5 9.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0 − − − − − − − − − − 1 log Age (yr) this work log M˙ acc (M yr− ) this work

Fig. 9.— Comparison of accretion properties estimated using stellar parameters of PMS stars in the LMC (from Spezzi et al. 2012). Left: A comparison of the ages estimated by Spezzi et al. (2012) using Tognelli et al. (2011) isochrones (ordinate) and in this work using Bressan et al. (2012) isochrones (abscissa) at the appropriate Z. Right: A comparison of M˙ acc derived using masses, radii using Bressan et al. (2012) isochrones (abscissa) versus those derived by Spezzi et al. (2012) (ordinate).

8.0 11 −

7.5 10 −

7.0

9 − 6.5

) from Spezzi et al.8 2012

6.0 1 − − yr

M 5.5 ( 7 − acc ˙ M log Age (yr) from Manara et al. 2012

5.0 log 6 −

4.5 6 7 8 9 10 11 4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0 − − − − − − 1 log Age (yr) this work log M˙ acc (M yr− ) this work

Fig. 10.— Comparison of accretion properties estimated using stellar parameters of PMS stars in the Orion Nebula (from Manara et al. 2012). Left: A comparison of the ages estimated by Manara et al. (2012) using Palla & Stahler (1999) isochrones (ordinate) and in this work using Bressan et al. (2012) isochrones (abscissa). Right: A comparison of M˙ acc derived using masses, radii from Bressan et al. (2012) isochrones, and LHα-Lacc relation used in this work (abscissa) versus those derived by Manara et al. (2012) (ordinate). The Lacc used to derive the M˙ acc was taken from Manara et al. (2012). )

PMS stars less than 1 Myr using diﬀerent isochrones. 1 − yr TESTING THE M˙ ESTIMATION METHOD 7.0 ACC M − We demonstrate that the method used to calculate ˙ ˙ 7.5 Macc reproduce the Macc estimated from a direct mea- − surement of the Hα line ﬂux in Fig. 11. We used the

Hα EW and stellar parameters (Teﬀ , M , R , AV , d) 8.0 of eight Herbig Ae/Be stars determined∗ by Pogodin∗ et − ˙ Pogodin et al. 2010 ( al. (2012), to calculate their Macc using the method de- 8.5 acc − ˙ ˙ scribed in Section 3.2 and 3.3. We compare the Macc cal- M

culated this way, with those estimated from the intensity log 8.5 8.0 7.5 7.0 − − − − 1 of Hα emission line measured from ﬂux-calibrated spec- log M˙ acc this paper (M yr− ) tra by Pogodin et al. (2012). We ﬁnd that our results ˙ accurately reproduce the measured M˙ within 0.15 dex, Fig. 11.— Comparison of Macc (abscissa) derived using the acc method followed in this paper versus traditional line diagnostic the error due to uncertainties in stellar parameters and methods (ordinate). All data taken from Pogodin et al. (2012). measured Hα EW.